How Using AI to Solve Problems May Make You Worse at Solving Them

New research has found something unexpected: using an AI assistant for just 10 minutes can make you noticeably worse at solving problems afterward, especially when you can no longer use the AI. Scientists from Carnegie Mellon University, MIT, Oxford, and UCLA studied hundreds of people doing math problems and reading tasks. When they took away the AI, people who had relied on it were more likely to give up or get answers wrong.

This pattern showed up across different types of problems, suggesting the AI's effect wasn't just about one skill but about how our brains approach problem-solving itself.

What Brain Scans Reveal

Researchers at MIT's Media Lab used brain scans to understand what's happening in your head when you use AI. They scanned people's brains while they wrote essays—some using ChatGPT, some using Google search, and some working alone. The scans showed that relying on AI changes how your brain works.

When people used AI, certain parts of their brains that normally handle thinking and remembering showed less activity. Researchers call this "cognitive debt." Think of it like outsourcing mental work to a tool: your brain gets lazy at the parts it's not using anymore.

Trust and Skill: A Tricky Balance

Other research looked at programmers using AI to solve coding problems. It found something counterintuitive: people who trusted AI the most often made worse choices about when to actually use it. They relied on the tool when they should have struggled through the problem themselves. But people with strong coding skills already in place were better at knowing when AI genuinely helped versus when it might hold them back.

How Students Actually Use AI

A survey of 1,000 college students showed they prefer having AI do the entire task for them rather than asking it for hints or guidance. This matches what the experiments showed: people lean on AI to finish work, not to improve their own abilities. Over time, this habit might interfere with learning, especially for students who already struggle with memory or focus.

Patterns We've Seen Before

History offers a useful comparison. When GPS navigation became widespread, studies found that people's ability to navigate and remember routes got worse. They could get places faster, but they lost some spatial skills. The AI research suggests something similar might be happening now—but with core thinking skills rather than navigation.

The broader question here is whether we gain something (immediate answers) while losing something else (the ability to figure things out ourselves). That trade-off is worth understanding, especially in workplaces where people still need to think independently and adapt to new situations.

What We Still Don't Know

The research comes with an important caveat: these studies used controlled settings with clear-cut tasks and short AI interactions. Real life is messier. People use AI across many areas of work and learning at the same time. We don't yet know if these short-term brain and behavior changes stick around permanently, recover over time, or grow worse with continued use.

What the research does show, plainly, is that AI assistance works both ways. It makes tasks easier right now, but it may make your brain less equipped to handle those same tasks on its own later. How organizations and schools manage that trade-off—whether they use AI in ways that build skills or replace them—will likely shape who can do complex work in the future.